Extract text from a PDF,free, in your browser

Drop one PDF onto the upload card, or click it to browse. Files can be up to 50 MB and there is no cap on the page count.

Click Extract text. Every page is read in turn and a counter shows which page it is on, so a long document does not look like it has hung. The text is joined together in reading order and shown to you before anything is saved.

The panel shows the character count, the page count, and the start of the text itself. That preview is the fastest way to tell a real text PDF from a scan: if the words are there, the extraction worked, and only then do you click Download .txt.

By default a line reading Page 1, Page 2 and so on is inserted where each page begins, which makes it easy to trace a quote back to its page. Untick the box for one continuous block instead. With the dividers off, pages that held no text are dropped completely rather than leaving a mystery gap.

A scanned PDF is a photograph of a document. There are no characters inside it, only pixels that look like characters, so there is nothing to extract and the tool says so instead of handing you an empty file.

The fix is text recognition. Run the scan through OCR to add a real text layer, then bring the result back here.

That is normal in a document that mixes typed pages with scanned inserts, such as a contract with a signed page photographed at the end. The tool counts the empty pages and tells you how many there were, so you know exactly which part is missing rather than assuming the whole file failed.

It does not. You get the reading order as plain text. Runs of spaces and tabs are collapsed to one space, so a table loses its alignment and a two-column page can interleave the columns, because the words come out in the order the PDF stored them rather than the order your eye reads them.

That is the right trade for feeding text into a search, a script, or a language model. When the layout is the point, convert to a document format instead.

Plain text is the wrong shape for numbers you want to sum. Rows come out as sentences and the column boundaries are gone. Use the spreadsheet converter, which keeps the rows and columns as cells you can sort and total.

A locked file cannot be read until the password comes off, so the upload card turns it away and points you at the unlock tool. Remove the password, extract from the plain copy, then put the password back on the original if you still need it.

Some PDFs carry a flag that tells a reader app to block selecting and copying. It is a request, not a lock, and it is not the same as a password. If the text is genuinely inside the file, it comes out here.

There is no page limit. A 500 page report works the same way as a two page letter, it just takes longer, and the page counter keeps you posted. The size ceiling is 50 MB per file, and a very long document can push a phone's memory, so a laptop is the better place for the big ones.

The .txt is written in a universal encoding, so accented Latin, Cyrillic, Greek, Arabic, Chinese and Japanese all survive exactly as they were in the PDF. Nothing is flattened to plain ASCII and nothing is replaced with a question mark.

Turn the page dividers off, download the .txt, and you have one clean block ready to index, diff, or paste into a prompt. Blank lines are tidied and repeated spaces are collapsed, which is usually exactly what a downstream tool wants.

If the reason you wanted the text was to find one clause or one number, extracting the whole document is the long way round. The question tool reads the PDF and answers directly, pointing at the page it came from.

The document is read inside the browser tab and the .txt is built there. Nothing is sent to a server, so a contract, a payslip or a medical record never leaves your machine, and there is no retention window to trust. There is also no hourly cap and no signup step. The free plan does count finished files, 10 a day shared across every tool.